A Study on Solar Power Generation Forecasting using Weighting Schemes based on Meteorological Data Correlation

Key Points

• This research aims to enhance the accuracy of solar power generation forecasts through weighted correlations with meteorological data.
• Analyzed correlations between solar power generation and various meteorological data using Pearson Correlation Coefficient (PCC).
• Applied six weighting methods to time series algorithms LSTM, CNN-LSTM, and CNN-LSTM-Attention.
• Compared performance against Random Forest algorithms to validate the effectiveness of deep learning models.
• LSTM utilizing linear weighting showed the highest prediction performance (R²: 98.8%) for ten types of meteorological data.
• Achieved approximately 3% improvement in performance compared to predictions without weighting.

Abstract